There is rising evidence of the health benefit associated with specific dietary interventions. Current food-disease databases focus on associations and treatment relationships but haven't provided a reasonable ass...
详细信息
With the rapiddevelopment of industry, the types and quantities of wastewater have increasedrapidly, and the pollution to water bodies has become increasingly widespread and serious. As far as water system is concer...
详细信息
Innovative methods to improve public transportation safety andresponse mechanisms have emergeddue to Internet of Things (IoT) and artificial intelligence (AI) technologies. IoT devices and Convolutional Neural Netwo...
详细信息
An alternative to human expert-performed manual identification is automatic detection of epilepsy using electroencephalogram (EEG) data. Automatic epilepsy detection from EEG data need high classification performance ...
详细信息
In the industrial sector, unplanned equipment failures can lead to significant financial losses, safety hazards, and operational inefficiencies. Traditional maintenance strategies, such as reactive maintenance (repair...
详细信息
ISBN:
(纸本)9798350354218
In the industrial sector, unplanned equipment failures can lead to significant financial losses, safety hazards, and operational inefficiencies. Traditional maintenance strategies, such as reactive maintenance (repair after failure) and preventive maintenance (scheduled maintenance), often prove inadequate in addressing these issues. This not only minimizes downtime but also optimizes maintenance resources and extends the lifespan of equipment. Machine learning models used in Predictive maintenance (PdM) are designed to analyze large volumes of data generated by industrial equipment. reinforcement learning optimizes maintenance schedules by learning from interactions with the environment, balancing the trade-off between maintenance costs and equipment reliability. Black-box models, such as deep learning, often provide high accuracy but lack transparency. Explainable AI (XAI) techniques are being developed to make these models more interpretable, enabling users to gain insights into the decision-making process and build confidence in the predictions. The deployment of ML models in industrial environments also requires seamless integration with existing systems. This involves addressing issues related to real-time data processing, scalability, and model retraining. Edge computing and cloud-based solutions are being explored to handle large-scale data processing and storage, while online learning techniques allow models to adapt to new data continuously. This paper presents several case studies from industries such as manufacturing, energy, transportation, and healthcare, illustrating the successful application of ML in PdM. In manufacturing, ML models have been used to predict failures in machinery such as turbines, compressors, and conveyor belts, leading to significant reductions in downtime and maintenance costs. The energy sector has employed ML for the maintenance of critical infrastructure, including power grids and wind turbines, enhancing operational efficienc
Wireless sensor networks (WSNs) consist of several nodes whose primary objectives are to monitor andregulate surroundings. Furthermore, sensor nodes are distributed in accordance with network utilization. The energy ...
详细信息
ISBN:
(数字)9798350368413
ISBN:
(纸本)9798350368420
Wireless sensor networks (WSNs) consist of several nodes whose primary objectives are to monitor andregulate surroundings. Furthermore, sensor nodes are distributed in accordance with network utilization. The energy consumption of sensor nodes is one of the most important problems in this kind of network. In fixed-sink networks, nodes in close proximity to the sink serve as intermediaries for transmitting data from other nodes to the sink. This results in a significant decrease in sensor energy usage. Consequently, the network’s lifespan decreases. due to their vulnerabilities, sensor nodes are prone to various attacks, including the vampire attack that poses a danger to WSN. This paper presents a novel approach for the detection and mitigation of vampire attacks using a Weighted Hopfield Neural Network (WHNN) combined with bio-inspired optimal path selection. The WHNN is employed to detect abnormal energy consumption patterns indicative of vampire attacks, leveraging its associative memory capabilities. To counteract the detected attacks, we incorporate a bio-inspired optimization algorithm, such as Aquila optimizer Algorithm (AOA), to dynamically select the most energy-efficient paths fordata transmission. This dual-layered strategy not only enhances the detection accuracy of vampire attacks but also ensures optimal path selection. due to their vulnerabilities, sensor nodes are prone to various attacks, including the vampire attack that poses a danger to WSN. The superiority of the WHNN-AOA technique is proved by simulation results, when compared to current schemes, based on performance measures such as packet delivery ratio, average throughput, detection ratio, and network lifespan, while also reducing the average residual energy.
As the automotive industry seeks to enhance environmental sustainability, predicting vehicle fuel efficiency has become crucial. This project aims to develop a deep learning model using TensorFlow to accurately predic...
详细信息
ISBN:
(数字)9798331501488
ISBN:
(纸本)9798331501495
As the automotive industry seeks to enhance environmental sustainability, predicting vehicle fuel efficiency has become crucial. This project aims to develop a deep learning model using TensorFlow to accurately predict fuel efficiency based on various factors such as vehicle specifications, driving conditions, and engine performance. A dataset sourced from the UCI Machine Learning repository was utilized, encompassing diverse vehicle types and their corresponding fuel efficiency metrics. Through data preprocessing and model training, our approach involves employing a neural network architecture optimized forregression tasks. The results demonstrate the model's ability to provide reliable predictions, contributing to more informeddecisions regarding vehicle performance and fuel management. Ultimately, this project serves as a stepping stone towards improving fuel efficiency in vehicles, promoting greener transportation solutions.
Object detection and pattern detection are fundamental problems in computer vision, and have real-world uses such as in autonomous vehicles, healthcare, and security. Over the recent years, improvements made in the de...
详细信息
ISBN:
(数字)9798331530013
ISBN:
(纸本)9798331530020
Object detection and pattern detection are fundamental problems in computer vision, and have real-world uses such as in autonomous vehicles, healthcare, and security. Over the recent years, improvements made in the deep learning framework have democratized these fields and made it easier forresearchers to build accurate models. CNN has become very popular in object detection and is currently implemented through features architectures like FasterrCNN and the Single Shot MultiBox detector (SSd). When it comes to pattern recognition, especially in sequential data, recurrent Neural Networks (rNN), especially Long Short- Term Memory (LSTM) network outperforms the rest. However, some of the important issues that remain unresolved include issues of robustness to varying environmental conditions and the question of computation cost. From the results and analysis of the research, it can be concluded that the following directions for future studies can be distinguished, enhance the model's robustness with methods such as domain adaptation, create new, less complex architectures forreal-time applications. Also, multi-modal fusion and unsupervised learning are the two areas with a brighter future in the field of automating machine learning. In sum, more advanced object detection and pattern recognition will continuously push the development and usability of these technologies for safer, more efficient, and intelligent systems at different domains.
This paper proposes a dual-input single-output (dISO) step-up dc-dc converter exhibiting high voltage gain for grid-connected applications. The main advantage of the proposed concept is the common ground between input...
详细信息
This paper proposes a dual-input single-output (dISO) step-up dc-dc converter exhibiting high voltage gain for grid-connected applications. The main advantage of the proposed concept is the common ground between input and output ports with a continuous input current profile at both input ports. The operating principle and steady-state analysis are presented in detail. The voltage gain expressions for both inputs are obtained using equivalent circuit analysis. The voltage and current stresses of all components and the design equations for L-C elements are derived. Preliminary simulation results are presented using the LTspice simulator to demonstrate the key operational characteristics of the proposeddISO step-up converter. Finally, the converter operation and analytical studies are verified through sample experimental results captured using a prototype converter.
How to learn discriminative video representation from unlabeled videos is challenging but crucial for video analysis. The latest attempts seek to learn a representation model by predicting the appearance contents in t...
How to learn discriminative video representation from unlabeled videos is challenging but crucial for video analysis. The latest attempts seek to learn a representation model by predicting the appearance contents in the maskedregions. However, simply masking andrecovering appearance contents may not be sufficient to model temporal clues as the appearance contents can be easily reconstructed from a single frame. To overcome this limitation, we present Masked Motion Encoding (MME), a new pretraining paradigm that reconstructs both appearance and motion information to explore temporal clues. In MME, we focus on addressing two critical challenges to improve the representation performance: 1) how to well represent the possible long-term motion across multiple frames; and 2) how to obtain fine-grained temporal clues from sparsely sampled videos. Motivated by the fact that human is able to recognize an action by tracking objects' position changes and shape changes, we propose to reconstruct a motion trajectory that represents these two kinds of change in the maskedregions. Besides, given the sparse video input, we enforce the model to reconstruct dense motion trajectories in both spatial and temporal dimensions. Pre-trained with our MME paradigm, the model is able to anticipate long-term and fine-grained motion details. Code is available at https://***/XinyuSun/MME.
暂无评论